[LU-11440] Make e2fsprogs-1.44.3-wc1 release Created: 27/Sep/18 Updated: 08/Dec/18 Resolved: 23/Oct/18 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.12.0 |
| Fix Version/s: | Lustre 2.12.0 |
| Type: | Improvement | Priority: | Major |
| Reporter: | Andreas Dilger | Assignee: | Minh Diep |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Issue Links: |
|
||||||||||||||||||||||||||||||||||||
| Rank (Obsolete): | 9223372036854775807 | ||||||||||||||||||||||||||||||||||||
| Description |
|
Need to make a release of e2fsprogs-1.44.3-wc1 from the master-lustre-test branch. This also needs a patch to lustre/Changelog to update the release version. |
| Comments |
| Comment by Andreas Dilger [ 02/Oct/18 ] |
|
Testing e2fsprogs-1.44.3-wc1 with SLES12 reports a filesystem corruption in conf-sanity test_99 "Adding meta_bg option" at mount time: LDISKFS-fs (dm-10): ldiskfs_check_descriptors: Block bitmap for group 0 overlaps block group descriptors LDISKFS-fs (dm-10): group descriptors corrupted! mkfs.lustre: Unable to mount /dev/mapper/ost1_flakey: Structure needs cleaning |
| Comment by Dongyang Li [ 05/Oct/18 ] |
|
Turns out the kernel of SLES12SP3 is missing a fix: 44de022c4382541cebdd6de4465d1f4f465ff1dd ext4: fix false negatives and false positives in ext4_check_descriptors() the bug was introduced in v4.4.140 and suse included the fix after v4.4.147, we are having the issue as the kernel is v4.4.143 on the test vms. I'm talking about the sles12 kernel versions here. What's our approach for this? open another Jira ticket and add this as a ldiskfs patch for kernels between v4.4.140 and v4.4.147 only? |
| Comment by Peter Jones [ 05/Oct/18 ] |
|
Dongyang If I understand correctly, will this disappear as soon as https://review.whamcloud.com/#/c/33236/ is landed (which may even be today)? Peter |
| Comment by Dongyang Li [ 06/Oct/18 ] |
|
Peter, Yes that's right, given we also use the new kernel on the test boxes. |
| Comment by Peter Jones [ 06/Oct/18 ] |
|
Then we should be all set now as it's landed |
| Comment by Andreas Dilger [ 13/Oct/18 ] |
|
It looks like SLES12 is now passing with e2fsprogs-1.44.3-wc1 so I think it makes sense to push master-lustre-test to master-lustre. Dongyang, can you please also push a patch to Lustre to update lustre/Changelog to reference the new version. |
| Comment by Gerrit Updater [ 15/Oct/18 ] |
|
Li Dongyang (dongyangli@ddn.com) uploaded a new patch: https://review.whamcloud.com/33370 |
| Comment by Dongyang Li [ 15/Oct/18 ] |
|
The old master-lustre branch is now v1.42.13.wc6-lustre master-lustre-test has been renamed to master-lustre lustre/Changelog also updated. |
| Comment by Andreas Dilger [ 15/Oct/18 ] |
|
Minh, the build artifacts are available for RHEL6/7, and SLES 11/12. Can you please please move them to the download site. If we don't already have a ticket for it, we also need to look into adding an Ubuntu builder for e2fsprogs. Building should be fairly straight forward, as the upstream Debian e2fsprogs is built directly from the e2fsprogs repo. |
| Comment by James A Simmons [ 17/Oct/18 ] |
|
The last testing with this gave: mke2fs 1.44.3.wc1 (23-July-2018) mke2fs: Size of device (0x448000000 blocks) /dev/mapper/crius-ddn-l12 too big to be expressed in 32 bits using a blocksize of 4096. It looks like a -o 64 option will be needed. I will look into it tomorrow to track down the issue. |
| Comment by Andreas Dilger [ 17/Oct/18 ] |
|
James, that must be formatting an MDT? By default the OST already uses "-o 64bit". With DoM it probably makes sense to enable "-o 64bit" on the MDT if it is large, but this forces the use of extents, which can hurt directory performance because directories are typically random 4KB blocks. |
| Comment by James A Simmons [ 17/Oct/18 ] |
|
Yes its the MDT being formatted that reports the problem. What should be the policy here? Give a warning and just format the disk at the maximum large size in the default case. BTW is the test harness now using this version of e2fsprogs? |
| Comment by Andreas Dilger [ 19/Oct/18 ] |
|
Yes, e2fsprogs-1.44.3.wc1 is currently being used by all autotest runs. I don't think the over-16TB MDT issue is related to e2fsprogs-1.44.3, so it should probably go into its own LU ticket. I do recall a similar issue being seen in the past, so if you search with the actual error message you may find it. |
| Comment by Gerrit Updater [ 23/Oct/18 ] |
|
Oleg Drokin (green@whamcloud.com) merged in patch https://review.whamcloud.com/33370/ |
| Comment by Peter Jones [ 23/Oct/18 ] |
|
Landed for 2.12 |
| Comment by Gerrit Updater [ 01/Dec/18 ] |
|
Andreas Dilger (adilger@whamcloud.com) uploaded a new patch: https://review.whamcloud.com/33766 |
| Comment by Nathaniel Clark [ 03/Dec/18 ] |
|
Ubuntu changes should be in one or more of the debian/control* files I believe. There doesn't seem to be an existing explicit dependency on e2fsprogs |
| Comment by Gerrit Updater [ 08/Dec/18 ] |
|
Oleg Drokin (green@whamcloud.com) merged in patch https://review.whamcloud.com/33766/ |